well suppose we have a communication channel and an sequence of
packets cross it. One records at the end of the channel whether one
received the packet or not as a 1010.... trace which is the population.

(each packet has a sequence number, packets cannot get reordered, so if 4
comes after 3, it means 2 is lost).

A receives a sample of size n1 corresponding to a certain set of packets.
B receives a sample of size n2 corresponding to *another* set of packets.

Assuming that when the packets are sent the characteristics of the channel
remain Exactly the same, the sample of A is independent of sample of B ?

Now, samples of A and B are from a bernoulli population. Can one use a
t-test to compare the difference of means of samples A and B ? (which is
used to compare the difference of means of samples from a normal
population)


vijay

-------------------------------------------

On Thu, 29 Apr 2004, Richard Ulrich wrote:

> On Thu, 29 Apr 2004 21:15:26 +0200, Vijay Arya <[EMAIL PROTECTED]>
> wrote:
>
> >
> > I guess tossing the same coin may not make it dependent. One can use a
> > single coin and generate an infinite trace of heads and tails 101001....
> > which becomes the population.
> >
> > Provided that A and B donot choose the same 0 or same 1 in their
> > samples, the samples of A and B will remain independent.
>
> When I toss a quarter multiple times and catch it in the air,
> the consecutive outcomes are not independent, or I never
> would have hit 16 heads in a row.
>
> Now, if it bounces on a hard surface ....
>
> --
> Rich Ulrich, [EMAIL PROTECTED]
> http://www.pitt.edu/~wpilib/index.html
>

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to